A Multilevel Modeling Approach Towards Wind Farm Aggregated Power Curve

Abstract

Wind turbine power curve modeling plays an important role in wind energy management. Accurate estimation of power curves can help reducing power systems maintenance costs. In this thesis, we use machine learning techniques such as clustering, spline regression as well as statistical learning approaches such as multilevel modeling and isotonic regression to reduce bias and/or variance of fitted power curves to improve their performance. First, we focus on reducing the effect of outliers in the wind speed-power data. To this end, we propose to enforce the inherent property of manufacturer power curve on fitted power curves to reduce outliers' impact. Manufacturer power curve is a worthy source of information about turbines' performance which has been ignored in the literature. So, we propose two nonparametric techniques based on the tilting method and monotonic spline regression methodology to preserve monotonicity on fitted power curves according to manufacturer power curve. Another challenging issue in fitting empirical power curve which was investigated seldom in the literature is heteroscedasticity of the wind speed-power data set. Age of turbine, location, air density, wind direction, and measurement errors are some of the reasons which may cause heteroscedasticity or non-homogeneity among observed data. To overcome this problem, we propose a novel methodology to use a hybrid estimation approach based on weighted balanced loss functions that account for both estimation error and goodness of fit by shrinking estimates toward standardized target models. Our proposed approach is very general and can be used with any desirable weighting scheme as an effective tool to improve the performance of existing power curve modeling approaches. Finally, we investigate improving wind farm aggregated power curve modeling instead of fitting power curves for individual turbines to reduce the complexity of wind farm management analyses. To solve this issue, we propose a novel clustering feature set based on turbines' overall performance and utilize K-Means clustering to classify turbines into homogeneous groups accordingly. We then apply multilevel modeling methods, including random intercept and random slope models on turbine clusters, to consider the hidden correlation among different clusters. We show that the proposed method is a solution for handling the complexity-accuracy trade-off issue since its accuracy is significantly higher than the single aggregated method alongside an equal complexity.

Dynamic Equivalent Modeling of a Large Renewable Power Plant Using a Data-Driven Degree of Similarity Method

“…[23] Wind Farm K-means, multilevel modeling Ref. [12] Stochastic security constrained Ref. [24] DFIG Fuzzy clustering Ref.…”

“…(3) Model 3: the equivalent model obtained by the single-unit method in Ref. [12]; (4) Model 4: the equivalent model obtained by the multi-unit method based on constraints of network structures in Ref. [12].…”

“…Dynamic equivalences of renewable power plants can be categorized into two distinct approaches: the single-unit and multi-unit methods [12]. The single-unit method entails the utilization of a solitary equivalent unit to represent the entire power plant, thereby streamlining the dynamic representation process.…”

Applications and Modeling Techniques of Wind Turbine Power Curve for Wind Farms—A Review

“…Meeting these high demands requires a massive penetration of "wind power" in the power grid system. However, the volatile nature of wind causes uncertainty and significant challenges in the "energy management systems" (EMS) in terms of scheduling and dispatching, which consequently impact the "reliability" of the power grid system [4][5][6]. This problem has drawn the attention of researchers towards the state of the art to develop appropriate solutions.…”

Turbine-level clustering for improved short-term wind power forecasting

“…A compromise between improving forecasting accuracy and maintaining a similar computational cost can be settled by grouping the turbines into homogeneous groups with data-driven clustering methods. For instance, reference [29] uses density peak clustering to group the turbines and consequently model the clusters as single entities, reference [30] groups the turbines using kmeans clustering to build the aggregated power curve of a wind farm, and reference [31] proposes a method based on fuzzy C-means clustering to classify the turbines with respect to the similarity of the wind speed-power curves.…”